I think we are still at the beginning of being able to truly exploit the potential of multicore architectures. There are simple ways of taking immediate advantage and more sophisticated approaches that are still being explored. One simple approach is to simply assign one operating system to one core and another to a second core. In this scenario, you can implement a real-time controller by running Windows on core A and some RTOS of choice on core B. Data and interrupts can be exchanged via the processor's communication mechanisms such as shared memory. In this case, the RTOS could reliably run a real-time control application and transfer data to the Windows side for operator monitoring. The operator could use all the spreadsheets, networking and user input without interfering with the real-time processes. This has already been implemented in a large number of systems.

Things get a bit more complicated when you want to spread a given application over multiple cores by, say, assigning specific threads to selected cores or even to optimize the performance of the previous scenario. Several companies have developed hypervisor technology to work with the hardware facilities, such as Intel's Virtualization Technology, to partition processor resources-core, memory, I/O devices-to create separate virtual computers, or to implement secure partitions that are safe from intruders but which allow traditional applications to run in other, less secure partitions. These efforts have resulted in definite gains in performance and versatility. For example, using Virtualization Technology along with Intel's Active Management Technology, it is possible to run a control program on a network of devices while at the same time setting up a virtual distributed application to develop on the same system without interfering with ongoing operations. This may sound redundant, but we are just beginning to realize some of the true potential of multicore architectures.

Submitted by Tom Williams on May 07, 2009